Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add skills/alebgl77/tierdecay/model-routingnpx skills add alebgl77/tierdecay --skill model-routinggit clone --depth 1 https://github.com/alebgl77/tierdecayWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/alebgl77/tierdecay/model-routing)<a href="https://agentmods.dev/skills/alebgl77/tierdecay/model-routing"><img src="https://agentmods.dev/badge/skills/alebgl77/tierdecay/model-routing.svg" alt="Measured on agentmods" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00072 | $0.00990 |
| Opus 5 | $0.00036 | $0.00495 |
| Sonnet 5 | $0.00014 | $0.00198 |
| Haiku 4.5 | $0.00007 | $0.00099 |
Grade A, and why
model-routing scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 3d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 87 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Model Routing
0. Ledger pre-check — posterior beats prior
Before scoring anything, read .claude/routing-ledger.md, in this order:
- Class has a live repo-playbook entry (not quarantined, floor not
reached) → dispatch one tier below the entry's provenance as a PROBE, entry
quoted in the brief (
tier-decay§3). A live entry outranks PRIORS — this is what keeps decay iterating. - Else task matches a class in PRIORS (≥3 logged rows) → use its empirical default tier, skip the rubric.
- Otherwise → score below. The rubric is only the cold-start prior.
1. Scoring rubric
Score the task on four axes; sum the points.
| Axis | 0 | 1 | 2 | 3 |
|---|---|---|---|---|
| Spec ambiguity | fully specified | minor gaps | requires design decisions | — |
| Reasoning depth | mechanical | standard logic | multi-constraint / algorithmic | novel algorithm, proof-like |
| Blast radius | 1 file | 2–5 files | >5 files or shared/core module | — |
| Risk surface | cosmetic | user-visible bug | data / perf / concurrency | security, money, irreversible |
Route:
- 0–3 → [T1]
executor(fast tier) - 4–6 → [T2]
heavy-executor(strong workhorse) - ≥7, or any axis at its maximum → [T3]: the frontier tier produces the spec
(main thread, or
oracleSOLVE) →heavy-executorimplements it →oracleREVIEWs the diff.
Tie-breaks: strong existing test coverage → down one tier. Critical path, or first task in an unfamiliar area → up one tier.
2. Fast lane (skip scoring when obvious)
- [T1] boilerplate, CRUD, tests of specified behavior, docs, renames, config, dependency bumps, bugfix with a clean repro.
- [T2] cross-module refactor, async/concurrency, migrations, perf work, gnarly bug without repro, implementing an already-designed API.
- [T3] architecture choices, novel algorithms, security-sensitive design, cross-cutting invariants — anything where being wrong is expensive to discover.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 3d ago First seen · 87 lines · 72 tokens per session scan A 9fe6585672da
model-routing is a skill published in the GitHub repository alebgl77/tierdecay (1 stars, last pushed 1mo ago), licensed MIT. It adds 72 tokens to every session and 990 once invoked, about $0.0004 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
cline-fix-volatile-msg
Ladder-aware Cline Anthropic caching — verify the rolling read/write ladder on the wire, then add the tools breakpoint and tune TTL. Updated for the 2026-08 AI-SDK monorepo.
continue-gemini-explicit
Continue's Gemini provider doesn't use the cachedContents API at all. Add explicit caching for sessions over the minimum token threshold.
aider-1h-ttl
Aider uses 5min TTL by default and works around long pauses with keepalive pings. Wire up the 1h TTL beta instead.
cline-openai-cache-key
Cline OpenAI native provider sends no promptcachekey. Add a stable per-task key so cachedtokens stops being zero.
continue-enable-defaults
Continue's prompt caching is opt-in via config and off by default. Flip the default to systemAndTools.
opencode-bedrock-doc-blocks
OpenCode places cachePoint on Bedrock messages containing DocumentBlocks, which produces a "nothing available to cache" error.